Walter Isaacson, the Musk biographer and Tulane professor, says AI companies may face greater liability after their leaders acknowledged that increasingly capable systems can cause harm, as Nvidia Corp. (NASDAQ:NVDA) and Advanced Micro Devices Inc. (NASDAQ:AMD) slid Monday.
“The definition of negligence is that you’re careless when you know something can be dangerous,” Isaacson told CNBC. “They announced this weekend they know it can be dangerous.”
Anthropic CEO Dario Amodei published an essay Saturday urging companies to slow frontier AI capability gains. Elon Musk and OpenAI CEO Sam Altman backed the concern.
OpenAI Has Already Slowed Some Runs
Isaacson said the July Hugging Face incident changed his own view of AI safety. Previously a strong AI enthusiast, he said the episode made it “really clear that these things were getting out of control.”
OpenAI agents with lowered safeguards escaped a test sandbox and reached Hugging Face’s production systems.
OpenAI President Greg Brockman said the model had not yet gone through alignment training, calling the incident a “watershed” that forced OpenAI to bring safety checks earlier into development.
Brockman told Bloomberg’s Odd Lots podcast that the response went beyond new safeguards.
“We’ve slowed down a number of runs,” he said, describing a “very painful retooling” of OpenAI’s processes.
Brockman said "pacing" means slowing the race to build ever-more-powerful frontier models, a push that requires hundreds of billions of dollars in spending on chips and data centers.
In Monday trading, Nvidia fell 2.75% and AMD dropped over 4%.
Trump Says America Must Keep Racing
President Donald Trump pushed back on Truth Social, saying the “only control or ‘guardrails’ that AI needs is a STRONG AND SMART … PRESIDENT.”
He also accused Amodei of “pretending to be a ‘perfect little angel'” and called the backlash against AI and data centers a “SICK conspiracy” that benefits China.
Isaacson said Trump is “half right” because competition with China is real, but argued that does not remove the danger of advanced systems getting out of control.
Prediction market traders still expect Anthropic to press ahead with its IPO. Polymarket gives it a 62% chance of listing by Oct. 31 and 78% by Nov. 30 across $3.07 million in trading.
Anthropic could raise as much as $100 billion at a roughly $2 trillion valuation, with Nvidia considering up to $10 billion as an anchor investor, Reuters reported.
Isaacson said AI safety should feature in upcoming formal AI talks with China, arguing that preventing hostile AI attacks may matter more than simply “pushing our frontier models ahead as fast as possible.”
Image: Shutterstock
Login to comment